Takamichi Toda

Learn More
Even as the sheer volume of information on the Web continues to expand, not everyone is able to find the information they seek using the conventional search methods available. Here we propose a solution in the form of a system that first assesses a user's circumstances and preferences in real time based on micromedia comments posted to the Web, then infers(More)
In our previous work, we developed AirTransNote, a student notesharing system that facilitates collaborative and interactive learning during regular lectures in conventional classrooms. Because taking notes on paper is a regular activity, our system does not impose an extra burden on students who share notes. However, in order to improve the effectiveness(More)
We generally use supervised learning when performing activity recognition using mobile sensor devices such as smartphones. In this application, case data associated with the sensor information and type of action is required. However, there is a possibility that a time shift occurs because this association is done manually on the audio and video that has(More)
In this paper, we propose a parallel distributed processing system for data-analytic project including human activity sensing flows, which manages dependency among data and analytic programs, and re-execute updated programs and dependent programs for updated data/programs. In the system, a data analyzer can specify the dependency and parts for requiring(More)
This study clarifies the characteristics of the time keys that are taken to enter Roman letters for Japanese sentences. First, test the appearance frequency of two consecutive characters of the alphabet in the input of English sentences and in the input of the Roman characters of Japanese sentences. Based on these results, we analyzed the features of stroke(More)
In this paper, we propose an approach to improve mobile activity recognition, given a training dataset with inaccurate segments, in which the beginning and ending timestamps of homogeneous and continuous activities have inaccurate boundaries due to human errors. In the proposed approach, we A) convert the training dataset to multilabel samples, B) train the(More)
  • 1